FITMYLLM · JULY 28, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K
Quant Bits VRAM @ 16K Quality Q3_K_M 4 14.2 GB
11.6 + 2.6 KV
low Q3_K_L 4.3 15.0 GB
12.4 + 2.6 KV
moderate IQ4_XS 4.46 15.5 GB
12.9 + 2.6 KV
moderate Q4_K_S 4.67 16.1 GB
13.4 + 2.6 KV
moderate Q4_K_M 4.89 16.7 GB
14.1 + 2.6 KV
good Q5_K_S 5.57 18.6 GB
15.9 + 2.6 KV
good Q5_K_M 5.7 18.9 GB
16.3 + 2.6 KV
good Q6_K 6.56 21.3 GB
18.7 + 2.6 KV
excellent Q8_0 8.5 26.7 GB
24.1 + 2.6 KV
lossless FP16 16 47.5 GB
44.9 + 2.6 KV
lossless
Select your GPU above to see speed estimates and compatibility for each quantization.
Deploying for a team or in production? Size GPUs, cost & scaling in Enterprise → ▸ READY TO RUN THIS? RENT BY THE HOUR
RENT A GPU AND RUN CODESTRAL 22B NOW
Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.
Community Ratings Chat Coding Reasoning Creative Vision Roleplay Agentic
Loading ratings...
Run this model Q3_K_M — 11.6 GB VRAM Q3_K_L — 12.4 GB VRAM IQ4_XS — 12.9 GB VRAM Q4_K_S — 13.4 GB VRAM Q4_K_M — 14.1 GB VRAM Q5_K_S — 15.9 GB VRAM Q5_K_M — 16.3 GB VRAM Q6_K — 18.7 GB VRAM Q8_0 — 24.1 GB VRAM FP16 — 44.9 GB VRAM
Ollama llama.cpp vLLM LM Studio KoboldCpp Jan Docker
▸ Easiest way to get started · Beginners
DOCS ↗ curl -fsSL https://ollama.com/install.sh | shCOPY
$ ollama run codestral:22b-v0.1-q4_K_MCOPY
Downloads and runs automatically. Add --verbose for speed stats.
▸ SETUP GUIDE >_
Auto-setup with fitmyllm CLI Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.
Auto-detect GPU Live tok/s in chat Speed benchmarks 9 inference engines
GPUs that can run this model At Q4_K_M quantization. Sorted by minimum VRAM.
Find the best GPU for Codestral 22B
Build Hardware for Codestral 22B Codestral — Mistral's dedicated coding model. 80+ programming languages.
Read full model card ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
Codestral 22B — 22.2B Dense. ▸ SPECIFICATIONS
PARAMETERS 22.2B
ARCHITECTURE Dense Transformer
CONTEXT LENGTH 32K tokens
CAPABILITIES coding
RELEASE DATE 2024-05-29
PROVIDER Mistral AI
FAMILY mistral ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY Q3_K_M 4 11.6 GB 88% Q3_K_L 4.3 12.4 GB 90% IQ4_XS 4.46 12.9 GB 92% Q4_K_S 4.67 13.4 GB 93% Q4_K_M 4.89 14.1 GB 94% Q5_K_S 5.57 15.9 GB 96% Q5_K_M 5.7 16.3 GB 96% Q6_K 6.56 18.7 GB 97% Q8_0 8.5 24.1 GB 100% FP16 16 44.9 GB 100%
§ 01 BENCHMARK SCORES
HumanEval 81.1
MMLU-PRO 49.4
MATH 35.6
IFEval 65.7
BBH 52.8
GPQA 18.6
MUSR 17.1
MBPP 78.2
BigCodeBench 52.0
§ 02 RUN COMMAND
Run Codestral 22B locally with Ollama — needs 14.1 GB VRAM at Q4_K_M:
$ ollama run codestral:22b
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback 